Advanced Lane Lines Finding -- Project 2¶
The Goal of this Project¶
The goal of this project is to write a software pipeline to identify the lane boundaries in a video from a front-facing camera on a car. The camera calibration images, test road images, and project videos are provided
Steps:¶
- Calibrate the camera
- Undistort the image
- Convert video into perspective view
- Explore Color/Gradient thresholds to convert image into binary with maximum lane information
- Detect lane lines in each video frame
- Determin the lane curvature
- Overlay back to the original image
In [1]:
# load the related packages
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import cv2
import glob
from __future__ import print_function
from ipywidgets import interact, interactive, fixed, interact_manual
import ipywidgets as widgets
from moviepy.editor import VideoFileClip,clips_array
from IPython.display import HTML
In [2]:
%matplotlib inline
#calculate objpoints and imgpoints
objpoints = []
imgpoints = []
#define the chess board corners
nx, ny = 9, 6
objp = np.zeros((9 * 6, 3), np.float32 )
objp[:, :2] = np.mgrid[0: nx, 0: ny].T.reshape(-1,2)
img_names = glob.glob('camera_cal/*.jpg')
fig, *axs = plt.subplots(4, 5, figsize = (20,10))
fig.tight_layout()
axs = np.ravel(axs)
for idx, img_name in enumerate(img_names):
img = mpimg.imread(img_name)
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
ret, corners = cv2.findChessboardCorners(gray, (nx, ny), None)
if ret == True:
objpoints.append(objp)
imgpoints.append(corners)
cv2.drawChessboardCorners(img, (nx, ny), corners, ret)
#cv2.imshow(img_name,img)
#cv2.waitKey(1000)
axs[idx].imshow(img)
axs[idx].axis('off')
'calibration1.jpg', 'calibration4.jpg' and 'calibration5.jpg' didn't show up because they don't show 4 * 9 corners
The following cell is to calculate ret, mtx, dist, rvecs, tvecs¶
In [3]:
import os
if not os.path.exists('output_videos'):
print('not exists')
os.mkdir('output_videos')
test_img = mpimg.imread( glob.glob('camera_cal/calibration2.jpg')[0])
img_size = (test_img.shape[1], test_img.shape[0])
ret, mtx, dist, rvecs, tvecs = cv2.calibrateCamera(objpoints, imgpoints, img_size, None, None)
undist = cv2.undistort(test_img, mtx, dist, None, mtx)
mpimg.imsave('output_images/calibrated2.jpg',undist)
fig, *axs = plt.subplots(2,2, figsize = (20,10))
axs = np.ravel(axs)
axs[0].imshow(test_img)
axs[0].axis('off')
axs[0].set_title('Original')
axs[1].imshow(undist)
axs[1].axis('off')
axs[1].set_title('Undistorted')
test_img = mpimg.imread( glob.glob('test_images/test6.jpg')[0])
img_size = (test_img.shape[1], test_img.shape[0])
undist = cv2.undistort(test_img, mtx, dist, None, mtx)
mpimg.imsave('output_images/calibrated_test6.jpg',undist)
axs = np.ravel(axs)
axs[2].imshow(test_img)
axs[2].axis('off')
axs[2].set_title('Original')
axs[3].imshow(undist)
axs[3].axis('off')
axs[3].set_title('Undistorted')
Out[3]:
In [4]:
#%matplotlib qt
test_img = mpimg.imread( glob.glob('test_images/straight_lines1.jpg')[0])
undist_img = cv2.undistort(test_img, mtx, dist, None, mtx)
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(15, 6))
#define src points and draw those points onto undistored image
src = np.float32([[205,720],[565,470],[720,470],[1100,720]])
A, B, C, D= tuple(src[0]), tuple(src[1]), tuple(src[2]), tuple(src[3])
undist_img_tmp = np.copy(undist_img)
cv2.line(undist_img_tmp, A, B, [255,0,0], 10)
cv2.line(undist_img_tmp, B, C, [255,0,0], 10)
cv2.line(undist_img_tmp, C, D, [255,0,0], 10)
cv2.line(undist_img_tmp, D, A, [255,0,0], 10)
mpimg.imsave('output_images/source_points_drawn.jpg',undist_img_tmp)
ax1.imshow(undist_img_tmp)
ax1.set_title('Undisorted Image with source points drawn', fontsize=15)
h, w = img.shape[:2]
#define dst points and draw those points onto warped image
dst = np.float32([[350, 720],[350, 0], [w-350, 0], [w-350, 720]])
A, B, C, D= tuple(dst[0]), tuple(dst[1]), tuple(dst[2]), tuple(dst[3])
img_size = (undist_img.shape[1], undist_img.shape[0])
M = cv2.getPerspectiveTransform(src, dst)
Minv = cv2.getPerspectiveTransform(dst, src)
warped = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
cv2.line(warped, A, B, [255,0,0], 10)
cv2.line(warped, B, C, [255,0,0], 10)
cv2.line(warped, C, D, [255,0,0], 10)
cv2.line(warped, D, A, [255,0,0], 10)
mpimg.imsave('output_images/dest_points_drawn.jpg',warped)
ax2.imshow(warped)
ax2.set_title('Warped image with dest. points drawn', fontsize=15)
Out[4]:
The following cell is to verify the perspective transform and generate a bird eye view video¶
In [5]:
test_imgs = glob.glob('test_images/*.jpg')
for idx, img_name in enumerate(test_imgs):
test_img = mpimg.imread(img_name)
undist_img= cv2.undistort(test_img, mtx, dist, None, mtx)
warped = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
mpimg.imsave('output_images/'+'undist_' +img_name.split('\\')[-1], undist_img)
mpimg.imsave('output_images/'+'warped_' +img_name.split('\\')[-1], warped)
fig, *axs = plt.subplots( 1,2, figsize=(10, 5))
axs = np.ravel(axs)
axs[0].imshow(undist_img)
axs[0].set_title('undisted '+img_name.split('/')[-1], fontsize=15)
axs[1].imshow(warped)
axs[1].set_title('warped '+img_name.split('/')[-1], fontsize=15)
In [6]:
def process_image(test_img):
undist_img= cv2.undistort(test_img, mtx, dist, None, mtx)
warped = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
return warped#np.concatenate((undist_img,warped),axis=-1)
#generate warped view of the video's
white_output = 'output_videos/warped_project_video.mp4'
clip1 = VideoFileClip("project_video.mp4")
white_clip = clip1.fl_image(process_image)
#final_clip = clips_array([clip1, white_clip])
%time white_clip.write_videofile(white_output, audio=False)
#%time final_clip.write_videofile('output_videos/warped_video.mp4',audio=False)
In [7]:
#combine the original video and warped view of the video together
clip1 = VideoFileClip("project_video.mp4")
clip2 = VideoFileClip('output_videos/warped_project_video.mp4')
final_clip = clips_array([[clip1, clip2]])
%time final_clip.write_videofile('output_videos/project_and_warped_video.mp4',audio=False)
In [8]:
white_output='output_videos/project_and_warped_video.mp4'
HTML("""
<video width="480" height="270" controls>
<source src="{0}">
</video>
""".format(white_output))
Out[8]:
Step 3.1: Color Space Exploring¶
RGB, HLS, HSV and LUVExploring RGB color space¶
- The following code is to check which channel in RGB color space can extract lane line information
- For bright pictures like test1.jpg or test4.jpg, or pictures with shadow like test5.jpg, it is hard to extract lane lines no matter it is red, green or blue channel
- Overall, the red channel gives more clear lane information
Observations of RGB color space:¶
| images | Red(215,255) | Green(195,255) | Blue(60,210) | Notes |
|---|---|---|---|---|
| straigh_line1.jpg | __$\checkmark$__ | __$\checkmark$__ | ||
| straigh_line2.jpg | __$\checkmark$__ | __$\checkmark$__ | ||
| test1.jpg | __$\checkmark$__ | bright picture | ||
| test2.jpg | __$\checkmark$__ | __$\checkmark$__ | ||
| test3.jpg | __$\checkmark$__ | __$\checkmark$__ | ||
| test4.jpg | __$\checkmark$__ | bright picture | ||
| test5.jpg | with shadow | |||
| test6.jpg | __$\checkmark$__ |
In [9]:
#convert image into red, green, blue, then pick different threshold
# for bright image like test1.jpg, thresh_R_low = 215
def RGB_split(thresh_R_low = 215,thresh_R_high = 255,\
thresh_G_low = 195,thresh_G_high = 255,\
thresh_B_low = 60, thresh_B_high = 210):
fig, *axs = plt.subplots(2*len(test_img_names),3,figsize=(10,5 * len(test_img_names)))
axs = np.ravel(axs).reshape(-1,3)
fig.tight_layout()
for i, img_name in enumerate(test_img_names):
name = img_name.split('_')[-1].split('.')[0]
test_imgs = mpimg.imread(img_name)
channel = test_imgs[:,:,0]
binary = np.zeros_like(channel)
binary[(channel>thresh_R_low)&(channel<thresh_R_high)]=1
axs[2*i][0].imshow(channel,cmap='gray')
axs[2*i][0].set_title(name+' Red',fontsize=10)
axs[2*i][0].axis('off')
axs[2*i+1][0].imshow(binary,cmap='gray')
axs[2*i+1][0].set_title(name+' Red Binary',fontsize=10)
axs[2*i+1][0].axis('off')
channel = test_imgs[:,:,1]
binary = np.zeros_like(channel)
binary[(channel>thresh_G_low)&(channel<thresh_G_high)]=1
axs[2*i][1].imshow(channel,cmap='gray')
axs[2*i][1].set_title(name+' Green',fontsize=10)
axs[2*i][1].axis('off')
axs[2*i+1][1].imshow(binary,cmap='gray')
axs[2*i+1][1].set_title(name+' Green Binary',fontsize=10)
axs[2*i+1][1].axis('off')
channel = test_imgs[:,:,2]
binary = np.zeros_like(channel)
binary[(channel>thresh_B_low)&(channel<thresh_B_high)]=1
axs[2*i][2].imshow(channel,cmap='gray')
axs[2*i][2].set_title(name+' Blue',fontsize=10)
axs[2*i][2].axis('off')
axs[2*i+1][2].imshow(binary,cmap='gray')
axs[2*i+1][2].set_title(name+' Blue Binary',fontsize=10)
axs[2*i+1][2].axis('off')
test_img_names = glob.glob('output_images/warped_test1.jpg')
interactive_plot = interactive(RGB_split,
thresh_R_low = (0,255,5),thresh_R_high = (0,255,5),\
thresh_G_low = (0,255,5),thresh_G_high = (0,255,5),\
thresh_B_low = (0,255,5),thresh_B_high = (0,255,5))
output = interactive_plot.children[-1]
#output.layout.height = '200px'
interactive_plot
Exploring HLS color space¶
- The following code is to check which channel in HLS color space can extract lane line information
- for bright pictures like test1.jpg or test4.jpg, the S channel gives much better lane line information
- for pictures with shadow like test5.jpg, HLS color space has no good solution
- overall, L channel gives better lane infor
Observations of HLS color space:¶
| images | H(14,85) | L(120,255) | S(210,255) | Notes |
|---|---|---|---|---|
| straigh_line1.jpg | __$\checkmark$__ | |||
| straigh_line2.jpg | __$\checkmark$__ | |||
| test1.jpg | __$\checkmark$__ | bright picture | ||
| test2.jpg | __$\checkmark$__ | |||
| test3.jpg | __$\checkmark$__ | |||
| test4.jpg | __$\checkmark$__ | bright picture | ||
| test5.jpg | with shadow | |||
| test6.jpg | __$\checkmark$__ |
In [10]:
#convert image into red, green, blue, then pick different threshold
def HLS_split(thresh_H_low = 14,thresh_H_high = 85,\
thresh_L_low = 120,thresh_L_high = 255,\
thresh_S_low = 210, thresh_S_high = 255):
fig, *axs = plt.subplots(2*len(test_img_names),3,figsize=(10,5 * len(test_img_names)))
axs = np.ravel(axs).reshape(-1,3)
fig.tight_layout()
for i, img_name in enumerate(test_img_names):
name = img_name.split('_')[-1].split('.')[0]
test_imgs = mpimg.imread(img_name)
#into HLS
hls = cv2.cvtColor(test_imgs, cv2.COLOR_RGB2HLS)
channel = hls[:,:,0]
binary = np.zeros_like(channel)
binary[(channel>thresh_H_low)&(channel<thresh_H_high)]=1
#binary[(channel>thresh_H_high)]=1
axs[2*i][0].imshow(channel,cmap='gray')
axs[2*i][0].set_title(name+' H',fontsize=10)
axs[2*i][0].axis('off')
axs[2*i+1][0].imshow(binary,cmap='gray')
axs[2*i+1][0].set_title(name+' H Binary',fontsize=10)
axs[2*i+1][0].axis('off')
channel = hls[:,:,1]
binary = np.zeros_like(channel)
binary[(channel>thresh_L_low)&(channel<thresh_L_high)]=1
axs[2*i][1].imshow(channel,cmap='gray')
axs[2*i][1].set_title(name+' L',fontsize=10)
axs[2*i][1].axis('off')
axs[2*i+1][1].imshow(binary,cmap='gray')
axs[2*i+1][1].set_title(name+' L Binary',fontsize=10)
axs[2*i+1][1].axis('off')
channel = hls[:,:,2]
binary = np.zeros_like(channel)
binary[(channel>thresh_S_low)&(channel<thresh_S_high)]=1
axs[2*i][2].imshow(channel,cmap='gray')
axs[2*i][2].set_title(name+' S',fontsize=10)
axs[2*i][2].axis('off')
axs[2*i+1][2].imshow(binary,cmap='gray')
axs[2*i+1][2].set_title(name+' S Binary',fontsize=10)
axs[2*i+1][2].axis('off')
test_img_names = glob.glob('output_images/warped_test1.jpg')
interactive_plot = interactive(HLS_split,
thresh_H_low = (0,120,2),thresh_H_high = (0,150,2),\
thresh_L_low = (0,255,5),thresh_L_high = (0,255,5),\
thresh_S_low = (0,255,5),thresh_S_high = (0,255,5))
output = interactive_plot.children[-1]
#output.layout.height = '400px'
interactive_plot
Exploring HSV color space¶
- The following code is to check which channel in HSV color space can extract lane line information
- for bright pictures like test1.jpg or test4.jpg, V channel gives better lane line information
- for pictures with shadow like test5.jpg, HSV color space has no good solution
- overall, V channel provides much bettern lane line information
Observations of HSV color space:¶
| images | H(14,85) | S(80,255) | V(220,255) | Notes | |
|---|---|---|---|---|---|
| straigh_line1.jpg | __$\checkmark$__ | ||||
| straigh_line2.jpg | __$\checkmark$__ | ||||
| test1.jpg | __$\checkmark$__ | bright picture | |||
| test2.jpg | __$\checkmark$__ | ||||
| test3.jpg | __$\checkmark$__ | ||||
| test4.jpg | __$\checkmark$__ | bright picture | |||
| test5.jpg | with shadow | ||||
| test6.jpg | __$\checkmark$__ |
In [11]:
#convert image into red, green, blue, then pick different threshold
def HSV_split(thresh_H_low = 14,thresh_H_high = 85,\
thresh_S_low = 80,thresh_S_high = 255,\
thresh_V_low = 220, thresh_V_high = 255):
fig, *axs = plt.subplots(2*len(test_img_names),3,figsize=(10,5 * len(test_img_names)))
axs = np.ravel(axs).reshape(-1,3)
fig.tight_layout()
for i, img_name in enumerate(test_img_names):
name = img_name.split('_')[-1].split('.')[0]
test_imgs = mpimg.imread(img_name)
#into HSV
hsv = cv2.cvtColor(test_imgs, cv2.COLOR_RGB2HSV)
channel = hsv[:,:,0]
binary = np.zeros_like(channel)
binary[(channel>thresh_H_low)&(channel<thresh_H_high)]=1
axs[2*i][0].imshow(channel,cmap='gray')
axs[2*i][0].set_title(name+' H',fontsize=15)
axs[2*i][0].axis('off')
axs[2*i+1][0].imshow(binary,cmap='gray')
axs[2*i+1][0].set_title(name+' H Binary',fontsize=15)
axs[2*i+1][0].axis('off')
channel = hsv[:,:,1]
binary = np.zeros_like(channel)
binary[(channel>thresh_S_low)&(channel<thresh_S_high)]=1
axs[2*i][1].imshow(channel,cmap='gray')
axs[2*i][1].set_title(name+' S',fontsize=15)
axs[2*i][1].axis('off')
axs[2*i+1][1].imshow(binary,cmap='gray')
axs[2*i+1][1].set_title(name+' S Binary',fontsize=15)
axs[2*i+1][1].axis('off')
channel = hsv[:,:,2]
binary = np.zeros_like(channel)
binary[(channel>thresh_V_low)&(channel<thresh_V_high)]=1
axs[2*i][2].imshow(channel,cmap='gray')
axs[2*i][2].set_title(name+' V',fontsize=15)
axs[2*i][2].axis('off')
axs[2*i+1][2].imshow(binary,cmap='gray')
axs[2*i+1][2].set_title(name+' V Binary',fontsize=15)
axs[2*i+1][2].axis('off')
test_img_names = glob.glob('output_images/warped_test1.jpg')
interactive_plot = interactive(HSV_split,
thresh_H_low = (0,255,2),thresh_H_high = (0,250,2),\
thresh_S_low = (0,255,5),thresh_S_high = (0,255,5),\
thresh_V_low = (0,255,5),thresh_V_high = (0,255,5))
output = interactive_plot.children[-1]
#output.layout.height = '400px'
interactive_plot
Step 3.2 gradient check¶
Exploring gradient in X, Y and Mag¶
- The following code is to check which direction of gradient can extract lane line information
- for bright pictures like test1.jpg or test4.jpg, gradient is hard to give better lane line information
- for pictures with shadow like test5.jpg, x-gradient gives much better lane line information
- overall, x-gradient provides much bettern lane line information
Observations of gradient:¶
| images | X(14,100) | Y(0,255) | ABS(0,255) | Notes | |
|---|---|---|---|---|---|
| straigh_line1.jpg | __$\checkmark$__ | ||||
| straigh_line2.jpg | __$\checkmark$__ | ||||
| test1.jpg | bright picture | ||||
| test2.jpg | __$\checkmark$__ | ||||
| test3.jpg | __$\checkmark$__ | ||||
| test4.jpg | bright picture | ||||
| test5.jpg | __$\checkmark$__ | with shadow | |||
| test6.jpg | __$\checkmark$__ |
In [12]:
#convert image into red, green, blue, then pick different threshold
def gradient_split(thresh_x_low = 14,thresh_x_high = 100,\
thresh_y_low = 80,thresh_y_high = 100,\
thresh_mag_low = 30, thresh_mag_high = 100,sobel_kernel=3):
fig, *axs = plt.subplots(len(test_img_names),4,figsize=(20,5 * len(test_img_names)))
axs = np.ravel(axs).reshape(-1,4)
fig.tight_layout()
for i, img_name in enumerate(test_img_names):
name = img_name.split('_')[-1].split('.')[0]
test_imgs = mpimg.imread(img_name)
#gray = cv2.cvtColor(test_imgs, cv2.COLOR_RGB2GRAY)
#gray = cv2.cvtColor(test_imgs, cv2.COLOR_RGB2HLS)[:,:,2]
gray = test_imgs[:,:,0]
sobel= cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
abs_sobel = np.absolute(sobel)
scaled_sobel = np.uint8(255 * abs_sobel / np.max(abs_sobel))
binary = np.zeros_like(scaled_sobel)
binary[(scaled_sobel>thresh_x_low)&(scaled_sobel<thresh_x_high)]=1
axs[i][0].imshow(gray,cmap='gray')
axs[i][0].set_title(name+' gray',fontsize=15)
axs[i][0].axis('off')
axs[i][1].imshow(binary,cmap='gray')
axs[i][1].set_title(name+' x gradient binary',fontsize=15)
axs[i][1].axis('off')
sobel= cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
abs_sobel = np.absolute(sobel)
scaled_sobel = np.uint8(255 * abs_sobel / np.max(abs_sobel))
binary = np.zeros_like(scaled_sobel)
binary[(scaled_sobel>thresh_y_low)&(scaled_sobel<thresh_y_high)]=1
axs[i][2].imshow(binary,cmap='gray')
axs[i][2].set_title(name+' y gradient binary',fontsize=15)
axs[i][2].axis('off')
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
# Calculate the gradient magnitude
gradmag = np.sqrt(sobelx**2 + sobely**2)
# Rescale to 8 bit
scale_factor = np.max(gradmag)/255
gradmag = (gradmag/scale_factor).astype(np.uint8)
# Create a binary image of ones where threshold is met, zeros otherwise
binary = np.zeros_like(gradmag)
binary[(gradmag >= thresh_mag_low) & (gradmag <= thresh_mag_high)] = 1
axs[i][3].imshow(binary,cmap='gray')
axs[i][3].set_title(name+' mag gradient binary',fontsize=15)
axs[i][3].axis('off')
test_img_names = glob.glob('output_images/warped_test5.jpg')
interactive_plot = interactive(gradient_split,
thresh_x_low = (0,255,2),thresh_x_high = (0,255,2),\
thresh_y_low = (0,255,5),thresh_y_high = (0,255,5),\
thresh_mag_low = (0,255,5),thresh_mag_high = (0,255,5), sobel_kernel=(3,9,2))
output = interactive_plot.children[-1]
#output.layout.height = '400px'
interactive_plot
Step 3.3 Generate binary image with lane information¶
- Red channel is chosen from RGB color space
- S channel is chosen from HLS color space
- S channel is chosen from HSV color space
- V channel is chosen from LUV color space
- X gradient is chosen
In [13]:
def BinaryPipeLine(raw_img,mtx,dist,M, final_only = False):
'''
raw_img: the original distorted image
mtx, dist: the matrix for camera distortion clibration
M: matrix to convert undistorted image into warp view
'''
img_size = (raw_img.shape[1], raw_img.shape[0])
#convert raw_img into undisted view
undist_img= cv2.undistort(raw_img, mtx, dist, None, mtx)
#genrate bird eye's view
warped_img = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
#RGB
sobel_kernel = 3
thresh_R_low = 215
thresh_R_high = 255
red_RGB = warped_img[:,:,0]
red_binary = np.zeros_like(red_RGB)
red_binary[(red_RGB>thresh_R_low)&(red_RGB<thresh_R_high)]=1
#HLS
thresh_S_low = 210
thresh_S_high = 255
hls = cv2.cvtColor(warped_img, cv2.COLOR_RGB2HLS)
s_hls = hls[:,:,2]
s_binary = np.zeros_like(s_hls)
s_binary[(s_hls>thresh_S_low)&(s_hls<thresh_S_high)]=1
#HSV
thresh_V_low = 220
thresh_V_high = 255
hsv = cv2.cvtColor(warped_img, cv2.COLOR_RGB2HSV)
v_hsv = hsv[:,:,2]
v_binary = np.zeros_like(v_hsv )
v_binary[(v_hsv>thresh_V_low)&(v_hsv<thresh_V_high)]=1
#LUV,
thresh_LUV_low = 155
thresh_LUV_high = 255
luv = cv2.cvtColor(warped_img, cv2.COLOR_RGB2LUV)
v_luv = luv[:,:,2]
luv_binary = np.zeros_like(v_luv )
luv_binary[(v_luv>thresh_LUV_low)&(v_hsv<thresh_LUV_high)]=1
#X,Y,and mag gradient
thresh_x_low = 14
thresh_x_high = 100
gray = warped_img[:,:,0]
sobel= cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
abs_sobel = np.absolute(sobel)
scaled_sobel = np.uint8(255 * abs_sobel / np.max(abs_sobel))
sobel_binary = np.zeros_like(scaled_sobel)
sobel_binary[(scaled_sobel>thresh_x_low)&(scaled_sobel<thresh_x_high)]=1
#combination of RGB, HLS, HSV, LUV and gradient into final binary image
final_binary = np.zeros_like(gray)
#final_binary[(red_binary==1) | (s_binary==1) | (v_binary==1) | (sobel_binary==1)] = 1
final_binary +=red_binary
final_binary +=s_binary
final_binary +=v_binary
final_binary +=luv_binary
final_binary +=sobel_binary
final_binary[final_binary<=1]=0
final_binary[final_binary>0]=1
if final_only:
return final_binary
return red_binary, s_binary, v_binary, sobel_binary, final_binary
imgs_name = glob.glob('test_images\*.jpg')
for i, img_name in enumerate(imgs_name):
test_img = mpimg.imread(img_name)
name = img_name.split('\\')[-1].split('.')[0]
red_binary, s_binary, v_binary, sobel_binary, final_binary = BinaryPipeLine(test_img,mtx,dist,M)
fig, *axs = plt.subplots(1,5,figsize=(20,5 ))
axs = np.ravel(axs)
fig.tight_layout()
axs[0].imshow(red_binary,cmap='gray')
axs[0].axis('off')
axs[0].set_title(name+' red binary', fontsize = 15)
mpimg.imsave('output_images\\'+name+'R_RGB.jpg',red_binary,cmap='gray')
axs[1].imshow(s_binary,cmap='gray')
axs[1].axis('off')
axs[1].set_title(name+' s binary', fontsize = 15)
mpimg.imsave('output_images\\'+name+'S_HLS.jpg',s_binary,cmap='gray')
axs[2].imshow(v_binary,cmap='gray')
axs[2].axis('off')
axs[2].set_title(name+' v binary', fontsize = 15)
mpimg.imsave('output_images\\'+name+'V_HSV.jpg',v_binary,cmap='gray')
axs[3].imshow(sobel_binary,cmap='gray')
axs[3].axis('off')
axs[3].set_title(name+' sobel_x binary', fontsize = 15)
mpimg.imsave('output_images\\'+name+'X_Gradient.jpg',sobel_binary,cmap='gray')
axs[4].imshow(final_binary,cmap='gray')
axs[4].axis('off')
axs[4].set_title(name+' final binary', fontsize = 15)
mpimg.imsave('output_images\\'+name+'_binary.jpg',final_binary,cmap='gray')
In [14]:
my_per_pix = 30/ 720
mx_per_pix = 3.7/ 700
def find_lane_pixels(binary_warped):
'''
this function is take bird's eye view image and find those pixels belong to left and righ lanes
inptus:
binary_warped is the bird's eye view of image in gray
outputs:
leftx, lefty: the coordinates of left lane pixels
rightx, righty: the coordinates of right lane pixels
out_img: the input image 'binary_warped' with lane pixels search rectanguler
'''
# Take a histogram of the bottom half of the image
histogram = np.sum(binary_warped[binary_warped.shape[0]//2:,:], axis=0)
# Create an output image to draw on and visualize the lane lines
out_img = np.dstack((binary_warped, binary_warped, binary_warped))
# Find the peak of the left and right halves of the histogram
# These will be the starting point for the left and right lines
midpoint = np.int(histogram.shape[0]//2)
leftx_base = np.argmax(histogram[:midpoint])
rightx_base = np.argmax(histogram[midpoint:]) + midpoint
# HYPERPARAMETERS
# Choose the number of sliding windows
nwindows = 9
# Set the width of the windows +/- margin
margin = 100
# Set minimum number of pixels found to recenter window
minpix = 50
# Set height of windows - based on nwindows above and image shape
window_height = np.int(binary_warped.shape[0]//nwindows)
# Identify the x and y positions of all nonzero pixels in the image
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Current positions to be updated later for each window in nwindows
leftx_current = leftx_base
rightx_current = rightx_base
# Create empty lists to receive left and right lane pixel indices
left_lane_inds = []
right_lane_inds = []
# Step through the windows one by one
for window in range(nwindows):
# Identify window boundaries in x and y (and right and left)
win_y_low = binary_warped.shape[0] - (window+1)*window_height
win_y_high = binary_warped.shape[0] - window*window_height
win_xleft_low = leftx_current - margin
win_xleft_high = leftx_current + margin
win_xright_low = rightx_current - margin
win_xright_high = rightx_current + margin
# Draw the windows on the visualization image
cv2.rectangle(out_img,(win_xleft_low,win_y_low),
(win_xleft_high,win_y_high),(0,255,0), 2)
cv2.rectangle(out_img,(win_xright_low,win_y_low),
(win_xright_high,win_y_high),(0,255,0), 2)
# Identify the nonzero pixels in x and y within the window #
good_left_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) &
(nonzerox >= win_xleft_low) & (nonzerox < win_xleft_high)).nonzero()[0]
good_right_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) &
(nonzerox >= win_xright_low) & (nonzerox < win_xright_high)).nonzero()[0]
# Append these indices to the lists
left_lane_inds.append(good_left_inds)
right_lane_inds.append(good_right_inds)
# If you found > minpix pixels, recenter next window on their mean position
if len(good_left_inds) > minpix:
leftx_current = np.int(np.mean(nonzerox[good_left_inds]))
if len(good_right_inds) > minpix:
rightx_current = np.int(np.mean(nonzerox[good_right_inds]))
# Concatenate the arrays of indices (previously was a list of lists of pixels)
try:
left_lane_inds = np.concatenate(left_lane_inds)
right_lane_inds = np.concatenate(right_lane_inds)
except ValueError:
# Avoids an error if the above is not implemented fully
pass
# Extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
return leftx, lefty, rightx, righty, out_img
def fit_polynomial(binary_warped):
'''
This function is to find the second order function to fit the left/right lane pixels
binary_warped is the bird's eye view of image in gray
return values:
pts_left: the left lane pixels
pts_right: the right lane pixels
out_img: the input image 'binary_warped' with lane pixels search rectanguler
curverad: the lane curvature values
'''
# Find our lane pixels first
leftx, lefty, rightx, righty, out_img = find_lane_pixels(binary_warped)
# Fit a second order polynomial to each using `np.polyfit`
left_fit = np.polyfit(lefty, leftx, 2)
right_fit = np.polyfit(righty, rightx, 2)
# Generate x and y values for plotting
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0] )
try:
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
except TypeError:
# Avoids an error if `left` and `right_fit` are still none or incorrect
print('The function failed to fit a line!')
left_fitx = 1*ploty**2 + 1*ploty
right_fitx = 1*ploty**2 + 1*ploty
## Visualization ##
# Colors in the left and right lane regions
out_img[lefty, leftx] = [255, 0, 0]
out_img[righty, rightx] = [0, 0, 255]
#curvature calculation:
left_fit_cr_0 = mx_per_pix / my_per_pix**2 * left_fit[0]
left_fit_cr_1 = mx_per_pix / my_per_pix * left_fit[1]
right_fit_cr_0 = mx_per_pix / my_per_pix**2 * right_fit[0]
right_fit_cr_1 = mx_per_pix / my_per_pix * right_fit[1]
y_eval = max(ploty)
left_curverad = ((1 + (2*left_fit_cr_0*y_eval*my_per_pix + left_fit_cr_1)**2)**1.5) / np.absolute(2*left_fit_cr_0)
right_curverad = ((1 + (2*right_fit_cr_0*y_eval*my_per_pix + right_fit_cr_1)**2)**1.5) / np.absolute(2*right_fit_cr_0)
curverad = np.average([left_curverad,right_curverad])
pts_left = np.array([np.transpose(np.vstack([left_fitx, ploty]))])
cv2.polylines(out_img, np.int32([pts_left]), isClosed=False, color=(255,255,0), thickness=3)
pts_right = np.array([np.flipud(np.transpose(np.vstack([right_fitx, ploty])))])
cv2.polylines(out_img, np.int32([pts_right]), isClosed=False, color=(255,255,0), thickness=3)
#pts = np.hstack((pts_left, pts_right))
#cv2.fillPoly(out_img, np.int_([pts]), (155,255,155))
return pts_left, pts_right, out_img, curverad
In [15]:
def draw_lane_lines_on_undist_img(raw_img):
'''
the functionis to draw lane lines to the undistored image
input:
raw_img: image before camera calibration
output:
result: image's with lane lines, this image is without camera calibration
'''
raw_img = raw_img[:,:,:3]
img_size = (raw_img.shape[0], raw_img.shape[1])
undist_img= cv2.undistort(raw_img, mtx, dist, None, mtx)
#warped_img = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
binary_warped = BinaryPipeLine(raw_img,mtx,dist,M, final_only = True)
#out_img = fit_polynomial(binary_warped)
pts_left, pts_right,_,_ = fit_polynomial(binary_warped)
lane_lines = np.zeros(img_size).astype(np.uint8)
lane_lines = np.dstack((lane_lines, lane_lines, lane_lines))
cv2.polylines(lane_lines, np.int32([pts_left]), isClosed=False, color=(255,255,0), thickness=3)
cv2.polylines(lane_lines, np.int32([pts_right]), isClosed=False, color=(255,255,0), thickness=3)
pts = np.hstack((pts_left, pts_right))
cv2.fillPoly(lane_lines, np.int_([pts]), (155,255,155))
img_size = (raw_img.shape[1], raw_img.shape[0])
lane_lines_unwarped = cv2.warpPerspective(lane_lines, Minv, img_size, flags=cv2.INTER_NEAREST)
result = cv2.addWeighted(undist_img, 1,lane_lines_unwarped, 0.2, 0)
return result
def draw_lane_lines_on_warped_img(raw_img):
'''
the functionis to draw lane lines to the bird's eye view image
input:
raw_img: image before camera calibration
output:
result: image's with lane lines, this image is bird's eye view version
'''
raw_img = raw_img[:,:,:3]
img_size = (raw_img.shape[1], raw_img.shape[0])
undist_img= cv2.undistort(raw_img, mtx, dist, None, mtx)
warped_img = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
binary_warped = BinaryPipeLine(raw_img,mtx,dist,M, final_only = True)
#out_img = fit_polynomial(binary_warped)
img_size = (raw_img.shape[0], raw_img.shape[1])
pts_left, pts_right,_,_ = fit_polynomial(binary_warped)
lane_lines = np.zeros(img_size).astype(np.uint8)
lane_lines = np.dstack((lane_lines, lane_lines, lane_lines))
cv2.polylines(lane_lines, np.int32([pts_left]), isClosed=False, color=(255,255,0), thickness=3)
cv2.polylines(lane_lines, np.int32([pts_right]), isClosed=False, color=(255,255,0), thickness=3)
pts = np.hstack((pts_left, pts_right))
cv2.fillPoly(lane_lines, np.int_([pts]), (155,255,155))
result = cv2.addWeighted(warped_img, 1,lane_lines, 0.2, 0)
return result
def draw_lane_lines_on_binary_warped_img(raw_img):
'''
the functionis to draw lane lines to the bird's eye view image
input:
raw_img: image before camera calibration
output:
result: image's with lane lines, this image is bird's eye view version in gray color
'''
raw_img = raw_img[:,:,:3]
img_size = (raw_img.shape[1], raw_img.shape[0])
undist_img= cv2.undistort(raw_img, mtx, dist, None, mtx)
warped_img = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
binary_warped = BinaryPipeLine(raw_img,mtx,dist,M, final_only = True)
#out_img = fit_polynomial(binary_warped)
img_size = (raw_img.shape[0], raw_img.shape[1])
_, _,out_img,_ = fit_polynomial(binary_warped)
return out_img
def combine_undist_warped_binary_img(raw_img):
raw_img = raw_img[:,:,:3]
img_size = (raw_img.shape[1], raw_img.shape[0])
undist_img= cv2.undistort(raw_img, mtx, dist, None, mtx)
warped_img = cv2.warpPerspective(undist_img, M, img_size, flags=cv2.INTER_LINEAR)
binary_warped = BinaryPipeLine(raw_img,mtx,dist,M, final_only = True)
#out_img = fit_polynomial(binary_warped)
img_size = (raw_img.shape[0], raw_img.shape[1])
pts_left, pts_right,binary_img,curverad = fit_polynomial(binary_warped)
#put curvature text
text = 'Lane Curvature: ' + '{:.3f}'.format(curverad) + 'm'
cv2.putText(undist_img,text,(undist_img.shape[1]//2-200, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (200,100,10), 2 )
lane_lines = np.zeros(img_size).astype(np.uint8)
lane_lines = np.dstack((lane_lines, lane_lines, lane_lines))
cv2.polylines(lane_lines, np.int32([pts_left]), isClosed=False, color=(255,255,0), thickness=3)
cv2.polylines(lane_lines, np.int32([pts_right]), isClosed=False, color=(255,255,0), thickness=3)
pts = np.hstack((pts_left, pts_right))
cv2.fillPoly(lane_lines, np.int_([pts]), (155,255,155))
warped_img = cv2.addWeighted(warped_img, 1,lane_lines, 0.2, 0)
img_size = (raw_img.shape[1], raw_img.shape[0])
lane_lines_unwarped = cv2.warpPerspective(lane_lines, Minv, img_size, flags=cv2.INTER_NEAREST)
undist_img = cv2.addWeighted(undist_img, 1,lane_lines_unwarped, 0.2, 0)
warped = cv2.resize(warped_img, (warped_img.shape[1]//3,warped_img.shape[0]//3))
binary = cv2.resize(binary_img, (binary_img.shape[1]//3,binary_img.shape[0]//3))
undist_img[0:warped.shape[0],0:warped.shape[1]] = warped
undist_img[0:binary.shape[0],(undist.shape[1] - binary.shape[1]):] = binary
return undist_img
In [16]:
# Load our image
raw_img= mpimg.imread('test_images/test3.jpg')[:,:,:3]
result = draw_lane_lines_on_undist_img(raw_img)
plt.imshow(result)
Out[16]:
In [17]:
# Load our image
raw_img= mpimg.imread('test_images/test3.jpg')[:,:,:3]
result = draw_lane_lines_on_warped_img(raw_img)
plt.imshow(result)
Out[17]:
In [18]:
# Load our image
raw_img= mpimg.imread('test_images/test6.jpg')[:,:,:3]
result = draw_lane_lines_on_binary_warped_img(raw_img)
plt.imshow(result)
mpimg.imsave('output_images/'+'curve_fit', result)
In [19]:
def undist_video_pipeline(img):
frame = np.copy(img)
result = draw_lane_lines_on_undist_img(frame)
return result
def warped_video_pipeline(img):
frame = np.copy(img)
result = draw_lane_lines_on_warped_img(frame)
return result
def binary_warped_video_pipeline(img):
frame = np.copy(img)
result = draw_lane_lines_on_binary_warped_img(frame)
return result
def merged_img_pipeline(img):
frame = np.copy(img)
result = combine_undist_warped_binary_img(frame)
return result
In [20]:
def undist_warped_binary_video(img):
frame = np.copy(img)
undist = draw_lane_lines_on_undist_img(frame)
warped = draw_lane_lines_on_warped_img(undist)
binary = draw_lane_lines_on_binary_warped_img(undist)
#print(warped.shape)
warped = cv2.resize(warped, (warped.shape[1]//3,warped.shape[0]//3))
binary = cv2.resize(binary, (binary.shape[1]//3,binary.shape[0]//3))
undist[0:warped.shape[0],0:warped.shape[1]] = warped
undist[0:binary.shape[0],(undist.shape[1] - binary.shape[1]):] = binary
return undist
In [21]:
out_img = mpimg.imread('test_images/test6.jpg')[:,:,:3]
result = merged_img_pipeline(out_img)
plt.imshow(result)
mpimg.imsave('output_images/'+'lane_area', result)
In [22]:
out_img = mpimg.imread('test_images/test6.jpg')
result = undist_video_pipeline(out_img)
plt.imshow(result)
Out[22]:
In [23]:
output = 'output_videos/undist_project_video.mp4'
clip1 = VideoFileClip('project_video.mp4')
clip = clip1.fl_image(undist_video_pipeline)
%time clip.write_videofile(output, audio=False)
clip1.reader.close()
clip.reader.close()
#video_clip.audio.reader.close_proc()
In [24]:
HTML("""
<video width="480" height="270" controls>
<source src='output_videos/undist_project_video.mp4'>
</video>
""")
Out[24]:
In [25]:
output = 'output_videos/warped_project_video.mp4'
clip1 = VideoFileClip('project_video.mp4')
clip = clip1.fl_image(warped_video_pipeline)
%time clip.write_videofile(output, audio=False)
clip1.reader.close()
clip.reader.close()
In [26]:
HTML("""
<video width="320" height="180" controls>
<source src='output_videos/warped_project_video.mp4'>
</video>
""")
Out[26]:
In [27]:
output = 'output_videos/binary_warped_project_video.mp4'
clip1 = VideoFileClip('project_video.mp4')
clip = clip1.fl_image(binary_warped_video_pipeline)
%time clip.write_videofile(output, audio=False)
clip1.reader.close()
clip.reader.close()
In [28]:
HTML("""
<video width="320" height="180" controls>
<source src='output_videos/binary_warped_project_video.mp4'>
</video>
""")
Out[28]:
In [29]:
output = 'output_videos/final_video.mp4'
clip1 = VideoFileClip('project_video.mp4')
clip = clip1.fl_image(merged_img_pipeline)
%time clip.write_videofile(output, audio=False)
clip1.reader.close()
clip.reader.close()
In [30]:
HTML("""
<video width="480" height="270" controls>
<source src='output_videos/final_video.mp4'>
</video>
""")
Out[30]: